ZINC-Curated is a high-quality, large-scale dataset of approximately 427 million drug-like molecules in SMILES format. To prioritize pharmaceutical relevance, it was constructed by curating a high-quality subset of the ZINC-22 database using a rigorous multi-stage filtration pipeline.
The pipeline enforces stringent physicochemical constraints, structural validity, medicinal chemistry rules, and diversity-aware stratification… See the full description on the dataset page: https://huggingface.co/datasets/SZU-ADDG/ZINC-Curated.
